Step 3: Search Quality
While building the search experience, the Admin should have assessed the search quality of the client’s top priorities queries based on their what kind of organization they are and what kind of questions they receive from their users.
As the reviewer, you can QA search quality based on several additional items:
Universal Prompt Results
All universal prompts should lead to very strong results. Click through each prompt to check. Ideally, each vertical shows up at least once through a universal prompt to ensure that all of the possible result types in the experience are showcased upfront.
Vertical Prompt Autocomplete Behavior and Results
Admins are encouraged to set up both static (hard-coded) and “mad-lib” style embedded fields vertical prompts for each vertical. It is best practice to enable at least
[[name]] as a vertical prompt for most verticals to create a more powerful predictive autocomplete experience. Try typing in various queries to ensure the autocomplete behavior is showing intuitive predictions that lead to expected results. If verticals have any hard-coded prompts and vertical prompts are
, click each prompt to ensure strong results are returned.
Verticals should be easily discoverable through all common synonyms matching the content type of that vertical. Example 1: a vertical for Acquisition FAQs should surface for queries like “merger”, “acquisition” and “transition”. Example 2: a vertical for financial professionals should generally surface for queries like “lenders”, “advisors”, and “officers”.
Location Search Quality (if applicable)
The location search algorithm can sometimes surface undesired or unexpected results, so it is important to test a few location-related searches. Try a few different types of queries here such as searching by city and by zip code to get a sense for how things are working, and check how the results are being sorted. Generally speaking, location entities are usually sorted by relevance to the query and then by proximity to the user.